THE PSYCHOTIC STATE Number 9 May 22, 1999 My Letter to Sun This month's article is a bit different. Psychotic as usual, I sent an e-mail - which summarized a conversation I had - to Jan LeVine, a sales rep for Sun. Sun is full of smart people, and I'd love to stir things up over there. As you will see, this letter is about Solaris and Linux. I first wrote about Free Software in the Psychotic State 8 months ago. Since these sorts of things that encourage religious wars, I don't want to write about them over and over. However, aside from being pschotic, I'm also lazy, and having written this I realized I didn't _also_ have to write another article. -------- Original Message -------- Subject: Regarding making Solaris free Date: Sat, 22 May 1999 18:52:15 -0400 From: Mitch Golden To: jan.levine@east.sun.com CC: mgolden@agency.com Dear Jan: It was a pleasure meeting you up in Newport, and I enjoyed our conversation. Apropos of our discussions, I'm sending you this letter about my opinion of what Sun ought to do with Solaris. If you really think that Scott McNealy will read this, I'd love it if you could forward it to him. If you think that there are refinements I should make to it before we send it along, I'd be happy to hear what they are. Let me outline what I think are the possible options for Sun with the Solaris OS: 1) Sell it as a proprietary OS tied to the Sun hardware. 2) Spin it off as a separate proprietary product, improved to the point that it runs just as well on the Sun hardware as on Intel (and possibly others). 3) Do Option 2, and give away the binaries, as closed source. 4) Make all versions of Solaris Open Source, inviting others to improve it. 5) Integrate the best features of Solaris into Linux, ultimately phasing out Solaris as a separate OS. Let me begin by saying that I believe that Sun is floundering right now, not really knowing how to deal with the (apparently) sudden emergence of Linux on the corporate scene. I think that this is because Sun somehow still, in its heart of hearts, can't acknowledge that the world has changed, and it can't get paid anymore for the time and effort that it put into the Solaris OS. A shame perhaps, but clear-eyed consideration of the way things are will lead you to realize that the only long-term solution is number (5). Consider (1). Every time I speak to Sun representatives, they tell me how much better Sun/Solaris is than Intel/WinNT. But Windows took over the world when it was in version 3.1, a crappy OS if ever there was one. Yes, I believe that Solaris is better than NT, but so what? It was far, far better than Win3.1, which nonetheless beat it hands down. There are lots of reasons for this, not least of which was the fact that Solaris was tied to a relatively expensive Sun platform, and that by being different from the other Unixes Sun kept the Unix market fragmented. Not enough applications were ported to Unix, so 3.1 won by default, despite its inferiority. So now WinNT isn't _quite_ as good as Solaris, but really, it's not as far off as it used to be. And there's no reason to assume that Microsoft won't narrow the gap further. In any event, nowadays you have an even more serious competitor than NT anyway: Linux itself. Lots and lots of applications are being ported. Whatever the advantages Solaris has over NT, the advantages over Linux are far, far less obvious. I was in my former career a theoretical physicist. When I left that field, in 1995, the most popular OS used was DEC's VMS. NeXTStep was also around a bit. Solaris was used on a few servers, but it was too expensive to be deployed on people's desktops. When I visit my physics friends nowadays, they are all, to a person, running Linux. I was an early adopter of Linux, but even I was surprised by the rapidity and uniformity of its adoption in the academic market. What this means is that Linux is poised to eat Sun's engineering workstation sales. These are the types of people who do not need wizards to set up their machines, who are not scared off by FUD. What is the value proposition to them of Solaris, when they will be able to run their fortran _and_ their word processors on Linux? Next let's consider option (2). I think I am qualified to comment here, because I am one of the few people - the only one I know, at any rate - who actually purchased a copy of Solaris X86. (The reason I did that was because at the time - two and a half years ago - my business couldn't afford a Sparc and Oracle didn't support Linux. That reason is now removed, of course.) The critical problem that Sun confronts here is that it regards Solaris as part of its competitive advantage. If Solaris ran as well on Intel as it does on Sparcs, Sun's hardware would have to compete on the basis of _price_ against Dell, Compaq, Gateway, et al. That's probably not an enticing prospect. As it currently stands, every Solaris X86 installation really represents one fewer Sparc sold (as it did in my case), not any gain in market penetration of Solaris vs. Windows NT. So Sun sells (or gives away) Solaris X86 secure in the knowledge that it will not really run as well on Intel as on Sparcs, and even if it does, the application support won't materialize. (Even Netscape gave up supporting Solaris X86 with its webserver, much to my chagrin.) So we see that the Solaris X86 effort is half-hearted at best. As it currently stands, it's really a waste of time. For the sake of argument, let's imagine that Sun were really serious. Suppose Sun totally spun off the Solaris OS into its own company. Would you invest in an enterprise that was attempting to challenge Microsoft (and Linux) in the OS market? Without the tie to the Sparc hardware, what would that OS offer? If you and I won't personally invest in SolariSoft, why should Sun? These considerations already dispense with Option (3). As I said above, each person who loads the free version of Solaris X86 on an Intel machine is really just one fewer Sparc sold, nothing more. Solaris is just not going to become the OS of choice on Intel hardware. Solaris X86 has no credible threat to Windows NT, and it doesn't have the mindshare of Linux. Making it free is not going to change that. For example, starting this week, Time Magazine (!) has a series about running Linux. (http://cgi.pathfinder.com/time/personal/19990524/tech.html) It's not credible to believe that Sun would get this sort of publicity for Solaris even if it gave it away. In any case, it's not the cost of Solaris X86 that keeps people from installing it, it's the lack of application support. If there were really the value there for installing Solaris on Intel hardware, people would gladly pay the few hundred dollars that Sun charges. Giving it away is not going to change that. Option (4) may seem seductive, but I believe it is doomed to failure as well. As Netscape learned with Mozilla.org, simply making a product Open Source does not guarantee that anyone will work for free to improve it. Bear in mind that one of the great paradoxes of building software - unlike virtually every other endeavor of humanity - is that people will work for free to do it. But obviously there's much, much more to getting people work for nothing than simply allowing them the right to do so. This is true in spades for Sun, since the positioning of Solaris is that it's Sun's operating system, not the public's. As someone who occasionally does work on Open Source software, I can tell you that I would be pretty loath to work for free just for the prospect of improving Sun's position against Hewlett-Packard. Even as non-proprietary a company as Red Hat has to watch very, very carefully to make sure it doesn't offend its free labor force. In any event, there are already lots of Open Source operating systems for Intel hardware, if people want to work on them. There are those who tell me that FreeBSD is better from a technical point of view than Linux. (I have no opinion on the matter.) A very few prefer the Gnu Herd. Whatever their merits, neither of these attracts the strong community of developers that Linux has. Why should Solaris, coming to the party last as it does? What is going to happen is that whatever is good about Solaris and is not in Linux is going to be incorporated into the latter. The Linux developers are going to do it no matter what happens, but they'll do it just a bit faster if the Solaris source is open. All told, that leaves us with Option (5). If you can't beat 'em, join 'em. Without a doubt there are some risks here, and changes Sun would have to make. If Sun pursues this course, it will lose its claim to proprietary software with which to sell its hardware. Any improvement in the operating system would immediately redound to the benefit of its competitors as well. However, what Sun gains is far greater than what Sun loses. It would have a rapidly evolving, state-of-the-art OS on which to base its solutions - at little cost to itself. It doesn't have to be the odd one out as companies finally port their solutions to Unix. If Sun joins the Linux bandwagon, what will result is that there will finally be a single standard for Unix. (In many people's view, including my own, the real reason that the Wintel duopoly took over is that the Unix vendors could never put aside their partisan differences and unite.) It appears that SGI is taking this path, integrating Irix with Linux (though for now only on Intel hardware). If Sun integrates Solaris with Linux, it would no longer spend opportunity cost on Solaris: Sun can free up many of the developers who now are working on the OS and use them to code other products which Sun can charge real money for, instead of trying to compete with a free OS. If Sun choses to integrate Solaris and Linux, it will have the best and cheapest possible product offering. In the long run, how could that not be the best course for Sun? Mitch Golden VP, Technology, New York AGENCY.COM The opinions expressed here are my own and are not necessarily those of AGENCY.COM.